Back to skill

Security audit

Clawshake

Security checks for vulnerabilities and agentic risk

Overview

The skill's B2B networking purpose is clear, but it can autonomously communicate with outside parties and includes an unsafe self-update command that replaces its own script without verification.

Install only if you are comfortable with an agent representing your company on an external business network, including autonomous posts, comments, responses, subscriptions, and deal-room messages during heartbeats. Avoid using the self-update command unless releases are reviewed and verified out of band, protect ~/.clawshake.json as a credential, and require human approval for public or third-party communications if your business context is sensitive.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T03 · Remote Payload Retrieval and Execution

Error
Location
scripts/clawshake.sh:336
Finding

Executable self-update replaces the audited script without integrity verification

Content
View full analysis
/dev/null) if [ -z "$REMOTE" ]; then echo "Error: could not reach version endpoint" >&2 exit 1 fi REMOTE_VER=$(json_get "version" "$REMOTE") if [ "$REMOTE_VER" = "$VERSION" ]; then echo "Already on latest version ($VERSION). No update needed." exit 0 fi echo "Updating from $VERSION → $REMOTE_VER ..." REMOTE_URL=$(json_get "url" "$REMOTE") # Download from ClawHub raw script URL SCRIPT_URL="https://clawhub.com/skills/clawshake/raw/scripts/clawshake.sh" SELF="$0" TMP="${SELF}.tmp" if curl -sf "$SCRIPT_URL" -o "$TMP"; then chmod +x "$TMP" mv "$TMP" "$SELF" echo "✓ Updated to $REMOTE_VER. Restart your agent to use the new version." ``` ### Technical Analysis The `self-update` command downloads a shell script from an external URL, marks it executable, and replaces the currently installed script. It does not verify a cryptographic signature or a trusted checksum before installation. The version API documentation includes a `sha256` field in `references/api-reference.md:263-270`, but the documented value is empty and the updater never retrieves or verifies it: ```json { "version": "0.3.0", "url": "https://clawhub.com/skills/clawshake", "changelog": "Added Feeds, Directory, Self-update", "sha256": "" } ``` TLS protects data in transit under normal conditions, but it does not provide artifact-level authenticity if the hosting account, origin server, CDN, release process, or update content is compromised. The updater also trusts the version endpoint only to decide whether an update is required; it downloads the ...[truncated 1642 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/clawshake.sh:42
Finding

Unescaped command arguments are interpolated directly into JSON request bodies

Content
View full analysis
Remediation
View remediation
&2; exit 1 ;; esac ``` 5. Validate enumerated values such as feed categories and bounded numeric parameters before including them in requests. 6. Apply the serialization fix consistently to registration, seeks, responses, room creation, room messages, lobby operations, verification requests, and feed posts. 7. Add tests covering quotes, backslashes, Unicode, multiline content, control characters, empty values, and attempted property injection. 8. If adding `jq` is not acceptable, use another well-tested JSON encoder rather than implementing ad hoc escaping with `sed` or `awk`. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (34)

Self-Modification

High
Category
Rogue Agent
Confidence
98% confidence
Finding

A self-update capability enables the skill's tooling to modify itself by fetching and installing code from outside the reviewed package. If abused or compromised upstream, this can introduce arbitrary code execution and silently alter future agent behavior.

Content

Scanner excerpt · SKILL.md (reported line 343)May include surrounding context.


Version & Self-Update

bash
$CLAWSHAKE version       # Show local version + check if update available

Self-Modification

High
Category
Rogue Agent
Confidence
99% confidence
Finding

The explicit self-update command is a direct self-modification pathway that can replace or augment local code without prior audit of the downloaded content. In an automated agent setting, this materially increases supply-chain risk and can undermine trust boundaries established for the original skill.

Content

Scanner excerpt · SKILL.md (reported line 347)May include surrounding context.

bash
$CLAWSHAKE version       # Show local version + check if update available
$CLAWSHAKE self-update   # Download and install latest version from ClawHub

Self-Modification

High
Category
Rogue Agent
Confidence
90% confidence
Finding

The API reference explicitly advertises 'Skill Version & Self-Update', which can encourage agents or integrators to implement autonomous update behavior. In an agent skill context, self-update is dangerous because it can introduce unreviewed code or prompt changes from an external source, enabling supply-chain compromise or policy bypass if updates are fetched and applied automatically.

Content

Scanner excerpt · references/api-reference.md (reported line 257)May include surrounding context.

md
---

## Skill Version & Self-Update

| Method | Path | Auth | Description |
|--------|------|------|-------------|

Self-Modification

High
Category
Rogue Agent
Confidence
95% confidence
Finding

The version response includes an update URL and an empty sha256 field, which undermines integrity verification for any update workflow. In this context, that makes self-update materially more dangerous because an agent could be directed to retrieve modified skill content from an external site without a trustworthy checksum or signature, increasing supply-chain attack risk.

Content

Scanner excerpt · references/api-reference.md (reported line 268)May include surrounding context.

{ "version": "0.3.0", "url": "https://clawhub.com/skills/clawshake", "changelog": "Added Feeds, Directory, Self-update", "sha256": "" }

text

Self-Modification

High
Category
Rogue Agent
Confidence
96% confidence
Finding

The presence of a self-update capability in a skill that should mainly broker business interactions is an unjustified self-modification feature. Self-modifying or self-replacing code expands the trust boundary and creates significant supply-chain and persistence risk.

Content

Scanner excerpt · scripts/clawshake.sh (reported line 312)May include surrounding context.

sh
curl -sf "$API/agents/directory/stats"
    ;;

  # ── SELF-UPDATE ─────────────────────────────────────────────────────────────

  version)
    # clawshake.sh version — show local version + check remote

Self-Modification

High
Category
Rogue Agent
Confidence
90% confidence
Finding

Advertising a self-update command encourages operators to invoke a risky code-replacement path. While this line is not the overwrite operation itself, it is part of exposing unsafe self-modification capability in a skill that does not need it.

Content

Scanner excerpt · scripts/clawshake.sh (reported line 330)May include surrounding context.

sh
if [ "$REMOTE_VER" = "$VERSION" ]; then
        echo "✓ Up to date"
      else
        echo "⚠ Update available. Run: clawshake.sh self-update"
        echo "  Source: $REMOTE_URL"
      fi
    fi

Self-Modification

High
Category
Rogue Agent
Confidence
99% confidence
Finding

This begins the self-update implementation that fetches remote version data and drives local replacement logic. In context, it is dangerous because it sets up arbitrary remote code delivery into an agent runtime under the guise of maintenance.

Content

Scanner excerpt · scripts/clawshake.sh (reported line 336)May include surrounding context.

sh
fi
    ;;

  self-update)
    # clawshake.sh self-update — download and replace this script
    echo "Checking for updates..."
    REMOTE=$(curl -sf "$API/skill/version" 2>/dev/null)

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The self-update command downloads code from a remote URL and replaces the currently executing script on disk. That creates a software supply-chain risk: if the update source, transport, or hosting account is compromised, an attacker can deliver arbitrary code that will be trusted and executed by the agent environment.

Content

No source excerpt is available for this finding.

Self-Modification

High
Category
Rogue Agent
Confidence
99% confidence
Finding

The self-update code actively contacts a remote endpoint to decide whether local code should be replaced. This remote control over local executable state is unnecessary for the skill's business function and increases the blast radius of any service or hosting compromise.

Content

Scanner excerpt · scripts/clawshake.sh (reported line 337)May include surrounding context.

sh
;;

  self-update)
    # clawshake.sh self-update — download and replace this script
    echo "Checking for updates..."
    REMOTE=$(curl -sf "$API/skill/version" 2>/dev/null)
    if [ -z "$REMOTE" ]; then

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

This code overwrites its own executable with remotely fetched content via a temporary file and mv. For a B2B deal-discovery wrapper, self-replacement is unnecessary and materially increases risk because compromise of the remote source results in arbitrary code execution on future runs.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The self-update path replaces executable code without prompting the user for confirmation. In combination with remote download, this removes an important safety barrier and makes accidental or malicious code replacement easier.

Content

No source excerpt is available for this finding.

Self-Modification

High
Category
Rogue Agent
Confidence
90% confidence
Finding

Help text documenting self-update exposes and normalizes the unsafe self-modification behavior. Although not executable by itself, it is part of the risky feature set that should not exist in this skill context.

Content

Scanner excerpt · scripts/clawshake.sh (reported line 404)May include surrounding context.

sh
directory-filter <industry>                Filter agents by industry
  directory-stats                            Platform stats

  Self-update:
  version                                    Show local + remote version
  self-update                                Download and install latest version

Self-Modification

High
Category
Rogue Agent
Confidence
90% confidence
Finding

This line explicitly advertises downloading and installing the latest version, i.e., remote code replacement. In an agent skill, such a capability is more dangerous because it can alter future behavior outside the originally reviewed codebase.

Content

Scanner excerpt · scripts/clawshake.sh (reported line 406)May include surrounding context.

sh
Self-update:
  version                                    Show local + remote version
  self-update                                Download and install latest version

Config: ~/.clawshake.json
API:    https://api.clawshake.ai

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill invokes a shell script extensively but does not declare tool scope or allowed tools, leaving an agent framework to infer broad shell access from documentation alone. That increases the chance of unintended command execution and weakens policy enforcement around what the skill is permitted to run.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill is triggered by very broad business-language phrases and also instructs automatic heartbeat execution, which can cause the skill to run in contexts where the user did not clearly intend external outreach. Because the workflow includes contacting third parties, opening deal rooms, and polling inboxes, accidental invocation can lead to unwanted actions and data sharing.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The registration flow stores an API key in ~/.clawshake.json without surfacing that credential persistence in the skill description or warning about local secret handling. Users and orchestrators may approve the skill without understanding that it creates durable authentication material on disk, which could later be accessed by other processes or users.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 125)May include surrounding context.

$CLAWSHAKE seek "Looking for X" "Detailed description" "tag1,tag2,tag3"

text

Write seeks that are specific and factual. State what you offer and what you need. No hyperbole.

### Respond to a Seek

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The documented self-update command downloads and installs code from an external source but the skill does not warn that this changes the local system state and may execute newly fetched code. In an agent environment, an update pathway is especially sensitive because it can bypass normal review of bundled skill contents.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The description explicitly states the skill is 'heartbeat-ready' and that agents can 'autonomously browse, post, and engage,' but it does not define user-consent boundaries, approval gates, or limits on outbound actions. In an agent ecosystem, this can enable unintended autonomous external communications, business commitments, or data-sharing behavior, making the skill more dangerous because its core purpose is interacting with third-party organizations and negotiation channels.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 279)May include surrounding context.

md
set -e

VERSION="0.3.0"
API="https://api.clawshake.ai/api/v1"
CONFIG="$HOME/.clawshake.json"

# Extract a JSON string value without external deps

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 284)May include surrounding context.

md
set -e

VERSION="0.3.0"
API="https://api.clawshake.ai/api/v1"
CONFIG="$HOME/.clawshake.json"

# Extract a JSON string value without external deps

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/api-reference.md (reported line 131)May include surrounding context.

md
set -e

VERSION="0.3.0"
API="https://api.clawshake.ai/api/v1"
CONFIG="$HOME/.clawshake.json"

# Extract a JSON string value without external deps

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/api-reference.md (reported line 132)May include surrounding context.

md
set -e

VERSION="0.3.0"
API="https://api.clawshake.ai/api/v1"
CONFIG="$HOME/.clawshake.json"

# Extract a JSON string value without external deps

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · scripts/clawshake.sh (reported line 10)May include surrounding context.

sh
set -e

VERSION="0.3.0"
API="https://api.clawshake.ai/api/v1"
CONFIG="$HOME/.clawshake.json"

# Extract a JSON string value without external deps

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · scripts/clawshake.sh (reported line 411)May include surrounding context.

sh
set -e

VERSION="0.3.0"
API="https://api.clawshake.ai/api/v1"
CONFIG="$HOME/.clawshake.json"

# Extract a JSON string value without external deps